is…

• the grande dame (first release in 1995)
• more than 10k packages on CRAN
• myriad builtin data sets
• vast statistical and numerical capabilities
• excellent (but sometimes opaque) documentation and
• standalone analyses

but

• can be slow and a memory glutton
• steep learning curve!

is…

• the Data Science debutante (although first release in 1991)
• explicitly object oriented
• encourages structured code
• general purpose and
• integration and production systems

but

• no native Data Science capabilities
• module dependencies can be a problem!

How can we
leverage
the
best
features of
both
languages?

What about blending them together?

Don’t think twice about mixing in a bit of SQL.

# R
#
dbGetQuery(db, "SELECT * FROM customer;")

# Python
#
db.execute("SELECT * FROM employee;")

Here are some options:

• \(\subset\) (R within Python)
• \(\subset\) (Python within R) or
• \(\{\) , \(\} \subset\) some other language.

Alternatives

• PypeR
  • release 1.1.2 (2014)
• pyRserve
  • uses Rserve (access R via RPC)
  • R process can run on separate machine
  • release 0.9.1 (2017)
• rpy2
  • runs embedded R in a Python process
  • release 2.9.4 (2018)

• Installs via pip.

pip install rpy2

• A Docker image is available.

docker run --rm -p 8888:8888 rpy2/jupyter

Alternatives

• rJython
  • release 0.0-4 (2012)
• rPython
  • release 0.0-6 (2015)
• SnakeCharmR
  • release 1.0.7 (2017)
• PythonInR
  • release 0.1-4 (2018)
• reticulate
  • runs embedded Python in a R process
  • release 1.8 (2018)

• Install from CRAN or GitHub.

# - From CRAN
install.packages("reticulate")
# - From GitHub
devtools::install_github("rstudio/reticulate")

Clearly
both
languages are
important.

A Data Scientist should be at least
conversant
in
both
of them.

Does it need to be an exclusive relationship?

We can
divide
our allegiance between
both
languages!